Race and the Race for the White House On Social Research in the Age of Trump - OSF
←
→
Page content transcription
If your browser does not render page correctly, please read the page content below
Race and the Race for the White House
On Social Research in the Age of Trump
Musa al-Gharbi1
______________________________________________________
Abstract: As it became clear that Donald Trump had a real base of political support, even as analysts
consistently underestimated his electoral prospects, they grew increasingly fascinated with the question of
who was supporting him (and why). However, researchers also tend to hold strong negative opinions about
Trump. Consequently, they have approached this research with uncharitable priors about the kind of person
who would support him and what they would be motivated by. Research design and data analysis often
seem to be oriented towards reinforcing those assumptions. This essay highlights the epistemological
consequences of these tendencies through a series of case studies featuring prominent and influential works
that purport to explain the role of race and racism in the 2016 U.S. presidential election. It demonstrates
that quality control systems, which should catch major errors, seem to be failing in systematic ways as a
result of shared priors and commitments between authors, reviewers and editors – which are also held in
common with the journalists and scholars citing and amplifying this work – leading to misinformation
cascades. Of course, motivated reasoning, confirmation bias, prejudicial study design, and failure to address
confounds are not limited to questions about Trump – however they seem to be particularly pronounced in
this case due to the relative homogeneity and intensity of scholars’ views about this topic as compared to
other social phenomena. “Trump studies,” therefore, provides fertile ground for exploring how social
research can go awry – and the consequences of these failures -- particularly with respect to work on
contentious and politically-charged topics.
Suggested citation: al-Gharbi, Musa (2018). “Race and the Race for the White House: On Social
Research in the Age of Trump.” The American Sociologist (49)4: 496-519.
1Paul F. Lazarsfeld Fellow in Sociology, Columbia University
Website: mussaalgharbi.com
Email: musaalgharbi@gmail.com
Some degree of bias is inevitable in social research. We are all shaped by our limited experience, our
personal commitments and aspirations, etc. – and these inform social inquiry in obvious and subtle ways
(Polanyi 1974). However, in a context where people with diverse and competing commitments,
backgrounds and experiences are simultaneously exploring a particular issue, these errors should approach
random distribution -- they would largely cancel each-other out. “Zones of agreement” should emerge in
the process, forming a basis for reliable knowledge (Shi et al. 2019). However, in a world where everyone
shares the same basic commitments, errors are not randomly distributed. They are far less likely to be
recognized as errors and addressed at all – indeed, they are likely to cascade as an integral part of the
“consensus” position, even giving rise to entire “null fields” of research (Ioannidis 2005).
Among social researchers, one commitment which virtually everyone seems to share is a passionate
distaste for Donald Trump, his agenda, and what he seems to represent. As it became clear that Trump was
a real contender for the Republican nomination, even as analysts consistently underestimated his prospects
-- both in the primary and the general election – researchers became transfixed by questions of just who
was supporting him and why (Samuelsohn 2016). These questions are perfectly reasonable and legitimate.
The problem is that, as a result of their disgust with Donald Trump, many social researchers also seem to
have developed strong assumptions about the type of person who would support him, and what they would
be motivated by. These presuppositions have consistently undermined research about the President and his
base. The most typical problems include:
• Failure to deal with obvious confounds
• Prejudicial study design
• Strange interpretations of data (apparently to make them fit with researchers’ preferred theses)
• Uncharitable interpretations of the words and actions of Trump and (especially) his supporters1 –
or even outright misrepresentation through cherry-picking and editing (Alexander 2016).
Here, we will present a series of case studies exploring prominent examples of this phenomenon -
- particularly as it relates to the role of race (and racism) in the 2016 U.S. presidential election. On this
point, an overwhelming consensus seems to have emerged that Trump voters were motivated largely (or
even primarily) by anti-minority or white-supremacist sentiments. It will be demonstrated that the evidence
scholars marshal in the service of these propositions is often weak, and abundant countervailing evidence
to the consensus position has been chronically under-addressed, while alternative explanations for Trump’s
victory have been insufficiently explored – despite apparently resting on surer empirical footing.
The works discussed in the following pages will be by scholars I respect and even admire. Indeed,
many are central to their respective fields. This is part of the point: the problem is not a matter of a couple
bad apples or fringe scholars – it cuts to the heart of who we are and what we do as social researchers. The
fact that the sort of errors described in the following pages could be made by people who obviously know
better – with the resultant works endorsed by so many others who also know better -- strongly suggests
motivated reasoning and confirmation bias are undermining the quality of our scholarship and the rigor of
our critiques (Koehler 1993, Mahoney 1977). Indeed, research has shown that the more important an issue
is perceived to be, the greater the risk that even grievous errors can go unnoticed or unaddressed (Wilson
et al. 1993). However, the election of Donald Trump is not just an ‘important issue’ -- for many of us, and
for many reasons, it is an outright crisis. However, this makes it all the more important to be clear-eyed
and level-headed about how we got here.
al-Gharbi, Musa (2018). “Race and the Race for the White House: On Social Research in the Age of Trump.”
The American Sociologist (49)4: 496-519.
p. 1
THEORETICAL CONSIDERATIONS
“The problem of knowledge is that there are many more books on birds written by ornithologists than books on birds
written by birds and books on ornithologists written by birds.” (Taleb 2010a, 108)
A key theme in the sociology of knowledge literature is that, contrary to idealized narratives of science as
a dispassionate enterprise in search of objective truths, science is in fact deeply political on several levels.
At the grand scale, ideological and cultural commitments inform who counts as trusted authorities and in
virtue of what, standards of appropriate evidence for a claim, the appropriate scope and most pressing
questions within a field of study, and even the structure of logical argumentation itself (al-Gharbi 2016a).
Expertise and scientific consensus are established within a field, and presented to outsiders, through
political processes internal to the discipline. At every stage in these processes there are power struggles
between rival coalitions (Latour 1988), as well as with experts in adjacent fields (e.g. economists, political
scientists, sociologists all aiming to shape policy on a particular issue, or competing for the same funding
opportunities, etc.). Throughout, there is a parallel negotiation between the expert class and the politicians
and publics they are supposed to advise and inform (Abbott 1988).
Yet another dimension of political contention is added for social scientific endeavors, because
unlike in the case of most material sciences, the objects of study in social research fields (i.e. people) have
agency. They have their own perspectives, interests, values, priorities and agendas which often diverge
sharply from those of researchers (Meadow 2013). Recent scholarship has held that not only should these
differences be recognized, ‘insider expertise’ should be privileged (Merton 1972). For instance, LGTBQ
people are understood as having some kind of special knowledge or authority on queer issues, women on
gender issues, racial and ethnic minorities (especially blacks) on race issues, Arabs and Muslims on the
Middle East and terrorism, etc. 2 This notion has undergirded much of the drive for diversity in social
research fields.
Curiously, despite the fact that there is far less political diversity in social research fields than there
is diversity in terms of gender, sexuality or even race (al-Gharbi 2018), the political positionality of scholars
seems to rarely be considered as important or problematic. While scholars are obliged to concede that there
is no superior race, class, gender, sexuality, or religion (except perhaps no religion),3 they are largely free
to take for granted that their own political and ethical commitments are correct, and that those who diverge
from them are both factually and morally ‘in the wrong.’ Indeed, while increased levels of education
correspond to reduced levels of racial and ethnic bigotry, they also correlate with significantly higher levels
of ideological prejudice (Henry & Napier 2017).4
This political blind spot could be problematic for any discipline. However, it has particularly strong
distortionary effects on the social sciences, because the subject of our fields is roughly identical to the
content of politics: sorting out how society is best arranged. The discipline of sociology was established in
the United States as a handmaid to the progressive reform movement (Abbott 2016, p. 49-50), and to this
day there remains a strong element of activism undergirding American sociology and most other social
research: scholars don’t want to merely understand social problems, but to help address them as well (Smith
2014).5 As a result of social researchers’ widespread certainty about the rightness of their political
convictions (Abbott 2018), and the desire to leverage our research and expertise to intervene in socio-
political disputes, it can be easy to temporarily lose sight of how complex, ambiguous and uncertain the
phenomena we study are.6 Consider:
al-Gharbi, Musa (2018). “Race and the Race for the White House: On Social Research in the Age of Trump.”
The American Sociologist (49)4: 496-519.
p. 2Even in “hard” disciplines such as physics, most experiments do not yield the expected result and
replications frequently fail (Lehrer 2009). As we approach fields like biomedical research, the replication
problem becomes far more severe – indeed, a majority of research efforts in these fields may be wasted by
avoidable errors and poor experimental design (Munafo et al. 2017). Yet, despite our best attempts to
emulate the material sciences (e.g. Levinovitz 2016), social research has not been able to demonstrate a
comparable level of reliability and power as these other fields (Horgan 2013). Indeed, there is far less
consensus on even foundational issues (Dunleavy, Bastow & Tinkler 2014). The effect-sizes of most
phenomena we study will be small almost by necessity (Gelman 2017), and it is likely that even many of
these effects will ultimately prove to be false-positives (Engber 2017). In other words, social researchers
should be cautious in making strong causal claims and skeptical when they encounter them from others.
Hence, there is a powerful a priori case for doubting deterministic narratives about the role of race
and racism in the 2016 general election. This served as the point of departure for the case studies that follow.
CASE STUDIES
Co-occurrence searches on Google Scholar7 can help highlight the scale of the phenomenon this essay seeks
to address. Restricting our analysis to 2016 and beyond, pairing “Donald Trump” with “racism” yields more
than 19.7k articles published in journals, books, or cited in academic work (Google Scholar). Searching
“Donald Trump” and “white supremacy” yields 5.9k results (Google Scholar) –“Donald Trump” and
“xenophobia” nets 6.6k (Google Scholar); “Donald Trump” and “ethnic nationalism” brings in 651 results
(Google Scholar). Even assuming some overlap between these, it is clear that an extensive literature has
emerged analyzing the relationship between Trump and racism, culled from a wide array of fields, and
produced in a relatively short amount of time (especially given how long it takes to publish academic
research, the lag time until works show up in Google Scholar after publication, and the restrictiveness of
the search criteria).
Having worked through the first ten pages of results for each of these (ten results per page), the
titles, descriptions and abstracts of these works suggest that many scholars take it more-or-less for granted
that Trump and his supporters are racist. A large number – even among those published in prominent
journals – seem to have an explicitly partisan and activist orientation. None of the hundreds of works
personally surveyed seemed to challenge prevailing narratives about Trump, his supporters, and their
alleged racism. This essay drills into a sample from this literature and addresses its major themes – focused
on works that have had demonstrably large impact both within the academy and beyond:
I. Ta-Nehisi Coates, “The First White President,” The Atlantic, October 2017.
It may seem strange to lead with Ta-Nehisi Coates in an article discussing problems with social research.
One anticipates a reply that runs basically, “so what if Ta-Nehisi’s research isn’t rigorous by the standards
of social science? That’s not his job. He’s a public intellectual, not an academic.” However, Coates’ work
greatly influences the academy, and informs a good deal of research in the humanities and social sciences.
Consider:
Studies suggest that a majority of journal articles are likely never read by anyone other than their
editors and authors (Eveleth 2014). 82% of articles in the humanities -- and roughly a third of all articles
published in the social sciences -- are never cited, even once (Remler 2014).
al-Gharbi, Musa (2018). “Race and the Race for the White House: On Social Research in the Age of Trump.”
The American Sociologist (49)4: 496-519.
p. 3These numbers, as bad as they seem, would actually look dramatically worse if self-citations were excluded
from consideration (Shema 2012). In contrast with this dismal state of affairs for most social research,
Coates’ work is widely cited in academic literature (Google Scholar). His memoir, Between the World and
Me, is among the most frequently-assigned common readings for incoming freshmen at universities
nationwide (Randall 2017), and his work is assigned as required reading for courses across a wide range of
disciplines (Google). He regularly participates on panels and delivers keynote addresses at universities, and
served as a visiting scholar at MIT from 2012 through 2014 (Smith 2017). He recently signed a three-year
contract to teach journalism courses at New York University – placing him in a pedagogical position at one
of the top-ranked journalism programs in the nation (Roy & Siu 2017), and at one of the most prestigious
universities in the world (Times Higher Education, 2017). In short, Ta-Nehisi Coates has already been far
more successful and influential in the academy than most ‘real’ academics could plausibly aspire to be over
the course of their careers.
Coates’ essay (2017b), “The First White President,” is far more likely to be read and cited by
academics – and to inform academic thinking on these issues – than most other instances of scholarship
that could be produced from the professional literature. It has already been cited at least 123 times (Google
Scholar) and was liked or shared more than 575.8k times on Facebook.8 Finally, Coates’ essay is highly-
representative of a swath of academic literature that’s out there – for instance, his central thesis is roughly
the same as that of Emory sociologist Carol Anderson’s (2017) book, White Rage: The Unspoken Truth of
our Racial Divide.9
Coates argues that while many pundits have pinned the election outcome on the ‘white working
class’ (e.g. Cohn 2016),10 this is an overly-narrow reading of the race. In fact, he asserts, Trump decisively
won among virtually all whites – cutting across income, gender, education, and geographic lines.11 From
there, Coates essentially imputes causation from correlation: (i)Trump repeatedly made racially/ culturally
inflammatory remarks; (ii) Trump won the white vote; (iii) Therefore, Trump won the white vote because
of his appeals to white supremacy. In his telling, white Americans rose up in 2016 to dismantle the legacy
of America’s first black president, Barack Obama (Coates 2017a) — and to reclaim what is rightfully
‘theirs.’ Trump was elected primarily because of racial resentment, and he maintains his hold on power by
playing to Americans’ latent sympathies with white supremacy. Despite the popularity of this thesis, it
suffers from several serious problems:
First, many of the white voters who proved most decisive for Trump’s victory actually voted for
Obama in both 2008 and 2012 (Uhrmacher, Schaul & Keating 2016). Indeed, according to one Democratic
political firm, the Global Strategy Group, these defections explain more than 70% of why Clinton lost
(Roarty 2017).12 If the white voters who ostensibly decided the 2016 election were horrified at the very
prospect of a black president, it is unclear why they would have supported Barack Obama’s initial
campaign. Similarly, if they were committed to undermining and dismantling the legacy of the first black
president, it is not clear why they would have voted to give him four years to further entrench his agenda
rather than simply voting for Mitt Romney in 2012. Apparently, instead of resisting Barack Obama in 2008
or 2012, they chose to act on their racial grievances years later, in a race between two white people. Coates
et al. portray the 2016 election as being fundamentally ‘about’ Barack Obama and what he allegedly
represents, rather than the (un)desirability of Hillary Clinton and/or Donald Trump. Yet in the lead-up to
the election, Obama was far more popular than both Clinton and Trump – indeed, he grew more popular
over the course of the 2016 cycle, even as the appeal of his would-be successors plummeted (Tesler 2016).13
All of this seems inconsistent with the notion that the election was an uprising against Barack Obama – as
a symbol or a politician.14
al-Gharbi, Musa (2018). “Race and the Race for the White House: On Social Research in the Age of Trump.”
The American Sociologist (49)4: 496-519.
p. 4Second, Trump did not mobilize or energize whites towards the ballot box (Mellnik et al. 2017).
Their participation rate was roughly equivalent to 2012 – and lower than in 2008. In fact, whites actually
made up a smaller share of the electorate than they had in previous cycles, while Hispanics, Asians and
racial “others” comprised larger shares than they did in 2012 or 2008. This seems inconsistent with the
thesis that Trump spearheaded an ethnic nationalist uprising.
Even if we overlook Obama’s continued popularity and whites’ tepid participation, at the very least
the ‘white supremacy’ thesis would seem to entail Trump winning a larger share of those whites who did
turn out to vote. In fact, Trump won a smaller share of the white vote than Mitt Romney did in 2012 (al-
Gharbi 2017b).15 However, he outperformed his predecessor with Hispanics and Asians, and won the
largest share of the black vote of any Republican since 2004. Given Trump’s lower share among whites, it
was likely these gains (and Democrats’ attrition) among people of color that put Trump in the White House.
Coates et al. therefore seem committed to arguing that the millions of these blacks, Latinos and Asians who
voted for Trump were also primarily or exclusively motivated by ‘white rage’ or their commitment to white
supremacy – or else conceding that it is possible to vote for Trump for other reasons (and of course, if this
is true of minorities, it stands to reason what whites could be similarly motivated by other factors).
Now, there might be some way to make the ‘white supremacy’ / ‘white rage’ argument work while
accounting for these apparent confounds – but this is not something that most proponents are doing.
Although the hypothesis is intended to explain voting behavior (i.e. why Trump won), researchers largely
ignore most of the critical data about who actually voted, and for whom. They run regressions testing
correlations between various demographic characteristics and political attitudes in pre-election surveys, but
fail to check if and how the resultant picture lines up with the available information on how people actually
voted. In those too-rare instances where researchers acknowledge apparent confounds for claims about
Trump voters’ alleged racism, the most common tactic is to say something along these lines:
“The explanation that Trump’s victory wasn’t an expression of support for racism because he got
fewer votes than Romney, or because Clinton failed to generate sufficient Democratic enthusiasm,
ignores the fact that Trump was a viable—even victorious—candidate while running racist
primary- and general-election campaigns. Had his racism been disqualifying, his candidacy would
have died in the primaries (Serwer 2017).”
There are three big problems with this line of thought:
First, there is a demographic confound. Again, Trump not only garnered stagnant turnout and a
smaller vote-share among whites relative to Mitt Romney – he was nonetheless able to win because he also
received a larger share among Asians and Hispanics (who did turn out in higher numbers), not to mention
African Americans. Should we assume that Asians, Latinos and blacks were also motivated primarily by
racism simply because they did not find Trump’s rhetoric (about China, Mexicans and Muslims)
disqualifying? This is an implication of the argument.
Second, it is just a logical error to assert that because Trump’s ‘racist’ rhetoric was not
disqualifying, Trump voters therefore supported him primarily due to his racial appeals. As Van Jones
aptly points out (Martinez 2017), Trump supporters regularly and forcefully rejected or expressed disquiet
with many of his racialized statements or policy proposals, but voted for him nonetheless because there
were other parts of his platform they found it essential to support, or because he was just viewed as the
least-worst option relative to Hillary Clinton. We can underscore this point with the help of a simple
“turnabout test” (Tetlock 1994):
al-Gharbi, Musa (2018). “Race and the Race for the White House: On Social Research in the Age of Trump.”
The American Sociologist (49)4: 496-519.
p. 5Hillary Clinton has a long record of foreign policy hawkishness – even surpassing most of the
Republican primary candidates in the 2016 election cycle (al-Gharbi 2015a). Her Democratic National
Convention was noted for its emphasis on patriotism, supporting the troops, confronting America’s
enemies, etc. to a point where many pundits observed that it actually seemed like a Republican event (e.g.
Gaouette 2016). Many prominent neoconservatives outright endorsed Clinton as the best candidate in either
party to promote their agenda (Norton 2016). Clinton actively sought out the endorsement of controversial
figures like Henry Kissinger and repeatedly bragged about their support on the campaign trail and in debates
(Grandin 2016).16 During the primaries, she relied on the same foreign policy consulting firm as hawkish
Republican candidates Marco Rubio and Ted Cruz (Fang 2015). Her general-election national security
advisory team was drawn heavily from people who oversaw some of the greatest excesses of the George
W. Bush administrations (Jilani, Emmons & LaChance 2016). Should we assert that because Clinton’s
well-established record of hawkishness was not disqualifying, that her supporters must have voted for her
because they, themselves, are hawks? Or to be less charitable: did most Democratic voters support Clinton
because they actively endorse the killing, traumatization and immiseration of others abroad (mostly people
of color, often of low socio-economic status) in the name of ill-defined ‘U.S. values and interests’? No. For
most Clinton voters this was an aspect of her record they found troubling, perhaps abhorrent, but not
disqualifying.17
Was race a critical factor in the 2016 election? Undoubtedly. But of course, this can be said of
virtually every electoral cycle in American history – including the 2008 and 2012 races which resulted in
the election and re-election of Barack Obama. The specific ways race influences politics are far more
complex and unpredictable than the analyses of Coates and his fellow travelers seem to suggest18 – even
among those who voted for Trump. Indeed, Hillary Clinton herself has a troubled history on racial issues
which many black voters did find disqualifying (al-Gharbi 2016b), perhaps helping to explain the depressed
turnout among this constituency – or even Trump’s gains among blacks (relative to Romney).
Finally, looking at whites’ stagnant turnout in 2016 and Republicans’ lowered vote share among
whites as compared to 2012 -- it could be that Trump’s inflammatory rhetoric was, on balance, a drag on
his candidacy among white voters, rather than being the key to his electoral success. In order to explore this
possibility, researchers would have to turn their attention away from the motivations of whites who did vote
for Trump, and towards the large majority (63%) of eligible non-Hispanic white voters who did not (Figure
1).
Figure 1
2016 U.S. Presidential Election
Voting Behavior Among Non-Hispanic Whites
Trump
36 37
Clinton
Other
Abstained
24
3
al-Gharbi, Musa (2018). “Race and the Race for the White House: On Social Research in the Age of Trump.”
The American Sociologist (49)4: 496-519.
p. 6For instance, even if it was assumed that 100% of those whites who voted for Trump were
motivated primarily by racial resentment – a claim which has no empirical basis to be clear – one would
still be unable to make sweeping generalizations about white people on the basis of these votes. Nearly as
many eligible non-Hispanic whites chose to abstain (36%) as those who voted for Trump (37%). This
should raise grave concerns about participation bias in interpreting what the election “means” on the basis
of Trump votes (Groves & Peycheva 2008, Paleologos et al. 2018) – especially given that an additional
27% of non-Hispanic whites explicitly rejected Trump by casting their votes for other candidates (Clinton,
Johnson, Stein, McMullin).
Another turnabout test can help underscore the problems with making sweeping claims on the basis
of such a dubiously representative sample: according to Bureau of Justice statistics, up to a third of all
African American males can expect to be incarcerated at some point in their lives (Bonczar 2003).
On a pie chart, the wedge for “black men expected to ‘do time’” would look very similar to the red area in
Figure 1. Should we assume on this basis that most black men are criminals? Of course not. Nor can we
declare an epidemic of racism among whites because 37% of eligible non-Hispanic white voters cast their
ballots for Trump (again, setting aside the grievous problems with imputing racist motives to those who did
support the Republican candidate simply on the basis of their vote).19
In short, the ‘white supremacy’ thesis not only lacks empirical support, it is confounded by the very
data it seeks to explain. Researchers often attempt to circumvent these realities by conflating (often
begrudging) tolerance for Trump’s racialized remarks and policies with active endorsement of his racial
appeals. Not only is this illogical, it is inconsistent with the way progressives seem to interpret support for
Hillary Clinton given her liabilities. Yet even working from within this model, it would actually be
consistent with the data to assert that most whites did find Trump’s rhetoric disqualifying. The ‘white
supremacy’ thesis therefore fails even if the conflation between tolerance and endorsement were allowed
to stand (although it should not be allowed to stand).
II. Thomas Wood, “Racism motivated Trump voters more than authoritarianism,”
Washington Post, 17 April 2017.
This article was written by an assistant professor of political science (Ohio State) and published in the
Washington Post “Monkey Cage” blog. All articles published on this site are curated and edited, not just
by academic political scientists, but by some of the most prominent scholars in the field (Sides 2013).
Therefore, although this essay was not a peer reviewed journal article, it was, in fact, reviewed by Wood’s
political science peers prior to publication at the Monkey Cage.20 The essay has had a demonstrable
academic impact – cited at least 31 times so far (Google Scholar), including in the American Sociological
Review and the British Journal of Sociology -- among the top journals in the field of sociology. The essay
has had a demonstrable public impact as well, with more than 71.2k shares on Facebook.
Wood seeks to challenge popular narratives about Trump voters being driven to the candidate due
to their support of authoritarianism (e.g. MacWilliams 2016). Utilizing data from the American National
Election Studies (ANES) Time Series Study, he found that Trump voters seem to be less supportive of
authoritarianism than those who voted for Romney or McCain (Figure 2).21
al-Gharbi, Musa (2018). “Race and the Race for the White House: On Social Research in the Age of Trump.”
The American Sociologist (49)4: 496-519.
p. 7Figure 2 (Wood 2017)
Wood then posits that, given Trump’s repeated use of racially inflammatory language throughout the
primary and general election, racism may better explain why a voter would support Trump over Clinton. In
order to test this hypothesis he developed a “symbolic racism” scale (SRS) and explored how white ANES
respondents’ scores on that scale corresponded to support for the Democratic and Republican presidential
candidates, from 1988 through 2016.
For the moment, we can set the well-known problems with attributing motives to political positions
on the basis of measures like “symbolic racism” scales (e.g. Sniderman & Tetlock 1986). 22
al-Gharbi, Musa (2018). “Race and the Race for the White House: On Social Research in the Age of Trump.”
The American Sociologist (49)4: 496-519.
p. 8And although both of these claims are contestable, we’ll just grant for the sake of argument that Dr. Wood’s
ANES-derived scales are sufficiently accurate measures of racist and authoritarian tendencies in the voters
sampled, and that his sample is sufficiently representative to allow for sound generalizations about Trump
and Clinton voters on the basis of his study. Figure 3 presents Wood’s visualization of his findings, as
presented in the Washington Post.
Figure 3 (Wood 2017)
One thing readers may immediately notice, which ends up being the centerpiece of his argument, is that the
gap between Democrats and Republicans on his symbolic racism measures was larger in 2016 than it had
been in previous years –
al-Gharbi, Musa (2018). “Race and the Race for the White House: On Social Research in the Age of Trump.”
The American Sociologist (49)4: 496-519.
p. 9suggesting that race likely played a much bigger role in the most recent election than it had in prior cycles.
This is an interesting finding which we’ll return to momentarily. In the meantime, however, it should be
noted that across all measures the longitudinal trend is roughly the same among supporters of the
Republican and Democratic candidates between 2004 and 2016: endorsement of ‘racist’ attitudes increased
substantially among all voters during the Obama years (2008, 2012) and then dropped among all voters in
this most recent cycle. That is, Wood’s data does not suggest that Trump voters were especially motivated
by racism. Instead, among supporters of both the Democratic and Republican candidates, whites were
actually less likely to endorse symbolically racist attitudes than they were in 2012. In other words, according
to Wood’s own data, whites who voted for Trump are perhaps less racist than those who voted for Romney.
The same holds among whites who voted for Clinton as compared to those who voted for Obama (in either
term).
This is truly a newsworthy and provocative finding about Trump voters: not only were they less
authoritarian than Romney voters, but less racist too! Had the findings been presented this way, Wood’s
report certainly would have served as a springboard for rich discussion and helped to push research on
Trump voters in new and interesting directions. Unfortunately, although his analysis of authoritarianism
focused on variation among Republicans over time, Wood curiously decided to adopt an alternative analysis
with respect to understanding the role of race23 – in particular, focusing on the gap between Democrats and
Republicans in 2016, in the service of a conclusion his data do not support.
Wood asserts, “Since 1988, we’ve never seen such a clear correspondence between vote choice and
racial perceptions.” He goes on to concede that, “The biggest movement was among those who voted for
the Democrat, who were far less likely to agree with attitudes coded as more racially biased.” This is an
understatement: across all measures, Republican voters’ SRS trend was actually towards convergence with
the other party. Change in Democrat-voters’ sentiment drives the entire divergence effect Wood refers
to…in an article that is supposed to explain Trump voters. Statistically speaking, the decline in
authoritarianism was greater than the decline in racism, while the decline in endorsing symbolic racism
among Democrats was larger than the decline among Republicans. So even through there were actually
declines across the board, in a multivariate analysis symbolic racism could be made to seem relatively more
significant for Republicans – if that was the image one was actively trying to convey. It seems bizarre and
incoherent to point to an apparent decline in endorsement of racism among Republicans in 2016 in order to
demonstrate that Trump voters are particularly racist – yet that is precisely what Wood did.
To elaborate: the gap between Democrats and Republicans was caused, not because Trump voters
were more likely to endorse symbolic racism than Republicans in previous cycles (again, the exact opposite
seems true), but instead because Democrats went out of their way to emphasize just how non-racist they
were in 2016 surveys — (ostensibly) rejecting symbolic racism more powerfully than they ever have in the
time 28 years of data presented. In fact, Wood’s data suggest that racist sentiment among Democrats
actually increased in 2008 with Barack Obama on the ballot, and more-or-less held at that higher level in
2012. This renders the unprecedented drop among Democrats in 2016 even more interesting. Indeed, if
Wood wanted to focus on the change in the partisan gap among SRS indicators, given that Democrats are
the ones driving that effect, the article should have been about them. Such a story would not suggest that
Trump voters are especially motivated by racism— again, Wood’s own data confound that claim. Instead,
perhaps Democratic voters were particularly motivated by anti-racism (McWhorter 2015).24 Wood could
have still said something like “moving from the 50th to the 25th percentile in the SRS scale rendered whites
20 percent more likely to vote for Clinton” – only now there wouldn’t be misleading insinuations about
Trump being made in the process.
al-Gharbi, Musa (2018). “Race and the Race for the White House: On Social Research in the Age of Trump.”
The American Sociologist (49)4: 496-519.
p. 10On the other hand, if Wood really wanted to make the article ‘about’ Trump voters, his results
should have been described in a completely different manner. For instance, the “unprecedented” partisan
divide among SRS indicators should not have been part of the story at all: given that it is entirely the product
of a severe longitudinal aberration among supporters of the Democratic candidate, it isn’t clear what (if
anything) can be inferred about Trump voters from the gap. With regards to the relative significance of
authoritarianism and racism, Wood could have more legitimately asserted that “Trump voters seem to be
more strongly influenced by racism than authoritarianism – although they were less motivated by either of
these factors than Republican voters have been in previous cycles.” This latter part, which Wood omitted,
is actually the significant story in his data on Trump voters (as compared to those who supported previous
Republican candidates).25 Wood should have also been more explicit about how much of the total variance
in voting behavior was explained by the factors he focused on (as reflected in the R2 of his regressions). It
seems likely that a majority of the variance would remain unexplained by his authoritarianism and/or racism
scales.
Following these standard best-practices would have had the added virtue of drawing readers’ minds
to the following question: “if both racism and authoritarianism seemed less appealing to Trump voters than
to previous Republican constituencies – and if these factors don’t explain most of the variance in voting
behavior anyway – is there some other (omitted) variable that may explain Trump voters even better?”26
This is important because, over and above the problems with Wood’s interpretation of his data, there is the
problem that the study design is prejudicial.
In this work, and others in the genre, it seems taken for granted that Trump voters must have been
motivated by some ‘bad’ factor (poverty, ignorance, irrationality, authoritarianism, racism, xenophobia,
sexism, etc.) and the task of social researchers is merely to identify which of these traits best explains the
data.27 Indeed, there is a long history in social research of pathologizing libertarians, conservatives and
populists – essentially describing views that diverge from researchers’ own political preferences as resulting
from ignorance, error or character defects (Duarte et al. 2014). Meanwhile, there is relatively little research
discussing the extent to which Clinton voters were motivated by negative impulses (the same ones or
others). It is implied that, by contrast, they must have been guided primarily by benign traits such as love
for others, patriotism, knowledge of the issues, rational self-interest, etc. (Stanovich 2017).
III. Arlie Hochschild, Strangers in Their Own Land: Anger and Mourning on the
American Right, September 2016.
Hochschild’s book was published two months prior to the 2016 election, and it is not technically
about that election. However, it is worth examining this work in closing for a few reasons. First, it had a
major impact of how people understood the outcome of the 2016 election. The perceived insight and
prescience of the work as it relates to the election of Donald Trump helped elevate it into being a New York
Times bestseller and National Book Award finalist. These same perceptions likely help explain why it is so
widely assigned as a required reading in course syllabi, in sociology departments and beyond (Google). It
has been cited more than 1.2k times (Google Scholar), often in works about the election. This is likely in
part because Hochschild herself has repeatedly asserted that Strangers in their Own Land robustly explains
Donald Trump’s appeal (Hochschild 2016b). In subsequent peer-reviewed articles she has implored other
social researchers to follow the methods she laid out in Strangers in their Own Land in order to gain further
insight into the “rise of the right” (Hochschild 2016c). Since the election, many social researchers and
journalists have attempted to do just this – flocking to regions Clinton lost, “on safari in Trump’s America”
(Ball 2017), in an attempt to figure out how Democrats can turn things around in 2018 and 2020.
al-Gharbi, Musa (2018). “Race and the Race for the White House: On Social Research in the Age of Trump.”
The American Sociologist (49)4: 496-519.
p. 11Given that we have so far explored the many ways statistical and historical facts have been distorted
to caricaturize Trump supporters as racist (not to mention ignorant, irrational, misogynistic, etc.),
ethnography may seem like a far more reliable method for understanding Trump voters and their
motivations. Researchers can directly observe the milieu of their subjects. Rather than the simple questions
pollsters rely on, ethnographers can have repeated in-depth conversations with their subjects – and follow
the discussion wherever it leads. Certainly, in many instances ethnographic research can promote key
theoretical insights or provide critical context to quantitative data (Wilson & Chadda 2009). However, there
is little reason to believe that ethnographic methods are, on balance, more reliable than quantitative or
historical approaches for understanding contentious political issues.
Contemporary research suggests that partisanship affects our cognition at a fundamental level –
potentially shaping our memories and evaluations of events, and at times, even coloring sensory perceptions
(Van Bavel & Pereira 2018). These are the very foundations upon which ethnography is built. Indeed, given
the highly subjective nature of the ethnographic encounter, the method may be even more vulnerable to
certain forms of bias, especially when researchers are seeking to understand perceived socio-political rivals,
with the aim of advancing a particular socio-political agenda (Avishai, Gerber & Randles 2013).28 Yet, even
as these commitments systematically distort or obscure critical details of their cases, ethnographers can
gain a strong erroneous conviction that they truly understand ‘the other side’ as a result of the intimate
nature of their research (Martin 2016). Seduced by the power of narrative, both the researcher and their
audience can significantly overestimate the reliability and generalizability of the resultant study (Taleb
2010b). This certainly seems to be the case with Strangers in Their Own Land.
Hochschild conducted forty “core interviews” (Hochschild 2016a, 247-50) -- all with white Tea
Party supporters, concentrated in one narrow region (Lake Charles) of a single state (Louisiana). This is an
extremely small and homogenous sample for explaining why some 63 million people, from all across the
country and all walks of life, cast their ballots for Donald Trump. Indeed, as we have previously explored,
the race was likely decided by people who voted for Obama in the previous election(s) but flipped for
Trump in 2016. Few of these voters would have been staunch Tea Party supporters (they may even have
very negative views about the Tea Party). And again, many of the most critical defections were not from
whites at all. Were it not for Republican gains (and Democratic losses) among blacks, Hispanics and Asians,
Trump likely would have lost -- given that he actually won a smaller share of the white vote than Mitt
Romney. Yet there does not seem to be one single county in the state Hochschild studied that flipped from
Obama to Trump (Figure 4).
al-Gharbi, Musa (2018). “Race and the Race for the White House: On Social Research in the Age of Trump.”
The American Sociologist (49)4: 496-519.
p. 12Figure 4 (Uhrmacher, Schaul & Keating 2016)
Indeed, by her own reckoning, only about 11% of white Louisiana voters supported Obama in 2012.
In 2011, up to 50% of Louisianans supported the Tea Party. The region is described by Hochschild herself
as the “geopolitical heart of the right” (p. 11-2). Simply put: these votes were not decisive in the 2016
election – there was no real contest for the Louisiana at all. And so, empirically speaking, it is not clear
how Hochschild’s project -- which basically attempts to understand why whites in a solidly-red state vote
Republican – would explain much about why so many former Obama supporters, many of them minorities,
voted for Donald Trump over Hillary Clinton in 2016. This is perhaps the most substantive question for
understanding why Trump won, and it is not addressed in her book – whatever its other merits. Yet, the
narrative Hochschild weaves is so seductive and compelling (especially for progressives) that questions of
representativeness and generalizability seem to be taken for granted – to include at times by Hochschild
herself – in discussions about the book and its relevance for understanding Trump’s appeal.
Moreover, while Hochschild’s subjects are often highly compelling and insightful when speaking
for and as themselves in the work, her analyses of these accounts frequently seem to suffer from political
bias. Hochschild’s treatment of her subjects is compassionate, but essentially condescending (Lozada
2016). Her main project is to understand the “emotional draw” of right-wing politics (Hochschild 2016a,
247), in part because she seems to take for granted that her subjects’ preferences could not be grounded in
a view of the world that is as logical or informed as that of progressives.29 Indeed, the book is fundamentally
structured around resolving the “great paradox” of why people who should support state programs,
interventions and regulations do not. That is, Hochschild is convinced from the outset that she understands
her subjects’ best-interests better than they, themselves, do. That assuredness never wanes – despite her
frequent criticisms of the ways other progressives demean and disrespect these constituents.
al-Gharbi, Musa (2018). “Race and the Race for the White House: On Social Research in the Age of Trump.”
The American Sociologist (49)4: 496-519.
p. 13With regards to the core contribution of the book, the ‘deep story’ of why her subjects support the
Tea Party (and Trump), Hochschild seems to put great weight on the fact that her respondents widely
endorsed the general picture she proposes – particularly the sense of frustration and marginalization that
comes through in the narrative. However, the fact that they support the broad picture does not necessarily
entail that they endorse every particular component of that narrative. Indeed, parts of the ‘deep stories’
Hochschild presents to her respondents seem to be intentionally leading. For instance (emphasis added):
“…You’re a compassionate person. But now you’ve been asked to extend your sympathy to all the
people who have cut in line ahead of you. And who’s supervising the line? It’s a black man whose
middle name is Hussein. He’s waving the line cutters on. He’s on their side.
He’s their president, not yours. What’s more, all the many things the federal government does to
help them don’t help you. Should the government really help anyone? Beyond that, from ahead in
line, you hear people calling you insulting names: ‘Crazy redneck!’ ‘White trash!’ ‘Ignorant
southern Bible-thumper!’ You don’t recognize yourself in how others see you.
You are a stranger in your own land. Who recognizes this?” (Hochschild 2016b, 686).
Every component of a “deep story” should be more-or-less essential. Yet evoking Obama’s race
and his middle name seems unnecessary here, an attempt to bait respondents into tacitly endorsing racially
inflammatory language. The implication seems to be that if they do claim to “recognize themselves” in the
general story, then they must have a problem with the specific fact that our president was “a black man,”
and one whose middle name happens to be “Hussein.” Does Hochschild really believe that her respondents
would find the bolded text integral to the story? Would the story have been less likely to resonate with them
if the relevant passage simply read: “…But now you’ve been asked to extend your sympathy to all the
people who have cut in line ahead of you. And the line supervisor seems to be just waving the cutters on.
He’s on their side...”? Would they spontaneously add, “and one more thing, the line supervisor is a negro
with a Muslim name!” This seems unlikely.
Indeed, if the “deep story” is supposed to explain why Louisianans support the right, then
presumably this tale would stretch beyond Barack Obama’s presidency in both temporal directions. The
geopolitical trends that left communities like Lake Charles behind began in the mid- 90’s. The oppositional
attitude towards the institutions, forces and actors perceived to be behind those trends also began to escalate
at this time (Foa 2016). Louisiana didn’t flip to the Republicans because of Obama – it has remained “red”
for every presidential election after 1996. Therefore, highlighting the fact that the “line supervisor” at the
time was a “black man whose middle name is Hussein” seems to substantially undermine the “depth” of
her narrative – reducing a story that has been ongoing for much of her subjects’ lives into something that is
fundamentally ‘about’ Barack Obama and what he allegedly represents. Similarly, Hochschild’s “deep
story” would presumably have remained just as true if Hillary Clinton had won in 2016 – which she almost
did -- therefore it seems strange to impute said “deep story” as explaining Trump’s victory.
At times, Hochschild’s project even seems exploitative: In a break with her groundbreaking
previous work, the goal of Strangers in their Own Land is not so much to understand these people on their
own terms, for their own sake – or to shine a light on the challenges they face, and their needs as they
understand them. Instead, Hochschild’s Tea Party supporting subjects and their stories are used as a means
to political ends -- ones which they would never, themselves, endorse (Ray 2017): Hochschild wants to
stall the “rise of the right”; she wants to help progressives understand how to advance their agenda and
candidates in red states; she wants to neutralize the appeal of Donald Trump.
al-Gharbi, Musa (2018). “Race and the Race for the White House: On Social Research in the Age of Trump.”
The American Sociologist (49)4: 496-519.
p. 14Hochschild seems to paternalistically believe that securing these political objectives would benefit her
subjects in the end – irrespective of their own preferences. However, this goal seems secondary at best in
reading the book. Although she typically refers to subjects as her “Tea Party friends” -- in the end they do
not seem to be on the same page, and they are not on the same team.
CONCLUSION
“…[T]he tendencies operating in 1948 electoral decisions not only were built up in the New Deal and Fair Deal era but
also dated back to parental and grandparental loyalties, to religious and ethnic cleavages of a past era, and to moribund
sectional and community conflicts. Thus in a very real sense any particular election is a composite of various elections and
various political and social events. People vote for a President on a given November day, but their choice is made not
simply on the basis of what happened in the preceding months or even four years; in 1948 some people were in effect
voting on the internationalism issue of 1940, others on the depression issues of 1932 and some, indeed, on the slavery
issues of 1860.” (Berelson, Lazarsfeld & McPhee 1986, 315-6).
The case studies presented here suggest serious, perhaps endemic, flaws on social research about the 2016
U.S. presidential election. Specifically, although the evidence presented here suggests that the role of race
has been widely overblown and misunderstood with respect to Trump’s victory, social researchers
nonetheless systematically represent Trump voters, and whites more broadly, as white supremacists in
virtue of their (presumed) support for the President. There is a Manichean subtext, with ‘evil’ racists
complicit in a depraved regime, v. the ‘good’ progressives who dutifully voted for Hillary Clinton (Hamid
2016). In the case of Hochschild, we see a variant on this narrative: Trump’s supporters aren’t bad people,
they’re just victims and hapless dupes of the Republican Party and their ‘fascist’ leader. It is easy to see
how these are edifying stories for progressives – which social researchers overwhelmingly are – but these
tendencies are also deeply pernicious in the following ways:
They threaten the viability of our discipline: Social science already faces a public perception
problem. Yet the troubled research discussed here targets the very people who are the most skeptical –
portraying them as racist (and often sexist, irrational and ignorant as well) on flimsy evidentiary grounds.
Yet the people being maligned in this research also happen to control the purse strings for federal research
funding, state funding for education, etc. (al-Gharbi 2017c). Their pundits already argue that social research
is shoddy leftist propaganda, scarcely worth serious engagement or attempted implementation. The works
discussed here provide ‘smoking guns’ to charges of bias. This is a dangerous state of affairs.
Cascade effects and feedback loops: When prominent and respected scholars publish flawed
research, it can pollute understanding of a subject matter in a way that becomes very difficult to correct.
Consider all the emerging scholars who have been assigned Hochschild or Coates’ recent works as a means
for understanding Trump’s election. Consider all of the subsequent research that will invariably reference
and attempt to build on these works, and the impact that this next generation of research will have in turn,
ad infinitum. In other words, the stakes of research about the last election extend well beyond the 2018 or
2020 races – myopic political scholarship can lead astray future activists and scholars in their own quests
to understand and address social problems.
Researchers can also help reify unfortunate social phenomena (Stampnitzky 2013, Watters 2011).
In this instance, unfairly labelling Trump voters as bigots -- simply because they voted for the Republican
in 2016 –
al-Gharbi, Musa (2018). “Race and the Race for the White House: On Social Research in the Age of Trump.”
The American Sociologist (49)4: 496-519.
p. 15this will likely stir up resentment against those who seem to be promulgating the charge (academics,
journalists and activists – especially minorities). Voters may double-down on their support for Trump, when
they might otherwise defect, simply in order to ‘stick it’ to those disparaging them (al-Gharbi 2015b). They
may even come to believe, on the basis of experts’ ubiquitous racialized narratives, that Trump is the best
candidate to promote white voters’ interests (Gray 2018) – despite whatever problems they may have with
what he has said or done so far. In other words, by demonizing Trump supporters as racist, social researchers
may be laying the groundwork for the very negative externalities they wish to avoid (i.e. Trump’s reelection,
increased resentment against the media, the academy, minorities, etc.).
In fact, one bitter irony of allowing research to be distorted by political aspirations is that the
unreliability of the resultant scholarship can undermine those very aspirations. In this instance, it is critical
for ‘the resistance’ to have an accurate understanding of who voted for Trump and why: inaccurate research
would likely generate ineffective electoral strategies. Indeed, motivated research and confirmation bias
likely explain why most analysts couldn’t see Trump’s strength in the last cycle (al-Gharbi 2017a). If
progressive scholars wish to avoid another humiliating and costly ‘black swan’ defeat for their party in
2020, then we’ll need to begin with a sober reckoning of how he won in the first place -- even if this means
being more charitable towards Trump voters than we feel naturally inclined to be, or being skeptical of
claims that would be very satisfying to accept.
Finally, it is normatively dangerous to countenance problematic research because it edifies our
identities. If scholars are sincerely worried about the way ‘Trumpism’ erodes and debases norms and
institutions we hold dear, we cannot effectively fight these trends by ourselves undermining the
methodological, evidentiary and disciplinary standards of our chosen professions. We should instead try to
defend and embody our values all the more, and to exemplify our best-practices – especially with research
relating to Trump and his supporters.
APPENDIX I:
John Sides, “Race Religion and Immigration in 2016,” Voter Study Group, June 2017.
In September 2017, prior to publication of the original version of this essay, I reached out to the editors of
the Monkey Cage (Washington Post) and expressed my concerns about the glaring problems in Wood’s
piece. They declined to issue a correction or retraction, nor to permit a rejoinder. Moreover, they
emphasized that even if Wood’s essay was flawed, his basic claim (Trump voters were motivated largely
by racism) was nonetheless correct—as evidenced by the wide literature of work in the field which comes
to similar conclusion without suffering the shortcomings I highlighted. To underscore this point they sent
along an essay by Vanderbilt political scientist John Sides (who also happens to be the publisher of the
Monkey Cage). Unfortunately, rather than ameliorating my concerns about the “Trump studies” literature,
the essay serves as yet another prime example of how the enterprise frequently goes awry.
Like Wood and Coates, Sides begins his research from the premise that because Trump repeatedly
used inflammatory language about minority and immigrant groups throughout the campaign, and
subsequently won the lion’s share of the white vote, these whites must have voted for Trump because of
his racialized rhetoric. His study purportedly demonstrates “cultural anxiety” to be the most explanatory
factor in explaining why white Americans voted for Trump over Clinton. 30 The proof offered for this claim
was that Voter Survey Group respondents’ attitudes toward immigration, blacks and Muslims better
predicted whether a participant voted Democrat or Republican in 2016 than they did in 2012. Figure 5
visualizes this data.
al-Gharbi, Musa (2018). “Race and the Race for the White House: On Social Research in the Age of Trump.”
The American Sociologist (49)4: 496-519.
p. 16You can also read
